Consciousness and how it got to be that way

Sunday, December 19, 2010

Epistemology and Encoding Rhetorical Intention in Language

Problems can arise when we misunderstand the rhetorical intentions surrounding informally stated hypotheses, which are certainly not limited to scientific endeavors. A stated hypothesis is a very strange kind of propositional utterance. When we state a hypothesis, we don't know whether it's true but we make the statement as if it were anyway - and we don't consider it to be lying or equivocating either. Why not? It has something to do with your audience knowing why you're uttering a proposition of uncertain truth value, which is exactly why if they don't know your proposition is a hypothesis, there can be problems. For example, Edward Sapir explicitly advanced the Penutian language family as a hypothesis to be investigated, and the theory was quickly adopted as gospel by the linguistic community, much to his distress. Fortunately it has largely been supported by data. It doesn't always work this way, in science or everyday conversation.

Since most languages mark questions and subjunctives explicitly, a fanciful solution would be a hypothesis particle. A simple phoneme would follow any informal statement of a hypothesis (-ba would work), and it would mean this: "I make no assertion about the truth of the proposition preceding this particle, but I want to learn the truth of this proposition, and I invite close scrutiny and criticism of this proposition in service of this goal."

But reflect further about rhetorical intention and truth value in utterances. Our classification of the rhetorical intentions is poor, since we don't recognize classes of utterances which are quite often explicitly encoded. A simple epistemological model of language is that all contentful utterances are commands, either directly to commit an action or to react to provided information, even if that's just for the listener to update his/her model of the world. As such we would expect most utterances will contain signals for rhetorical intention above and beyond the content of the sentence; there is the proposition being uttered, and the intention the utterer has of how the audience should react. Analytic philosophers attempted to approach natural language propositionally although their conclusions were sometimes provincial, hobbled as they sometimes were by an impoverished knowledge of the variety of language structures that existed outside Western Indo-European. Here is a phylogeny of coherent utterances which includes rhetorical intention.


1) COMMAND: "Get out of my house."

Extra-contential rhetorical tag: none.

1.1) Commands are the basic form of language and it is therefore not surprising that the command forms of verbs are usually morphosyntactically as, or more simple than, even the infinitive.


2) CONTENTFUL EXCLAMATION: "A blue hummingbird!"

Extra-contential rhetorical tag: "Recognize this object/event I have verbally pointed to."

2.1) Some languages (e.g. Tagalog, Washoe) do encode this intention explicitly and have focus markers which explicitly declare what the speaker wishes the listener to focus on. These markers occur throughout sentence structures and are not limited to noun-phrase exclamations like the one above.

2.2) Dependent phrase structure is always just another example of recursive phrase structure. That said, shorthand often evolves for parsimony (i.e., "There is a red book on the table" is really just shorthand for, and content-wise exactly the same as, "There is a book that is red on the table".) In languages with copulae like English, superficially there seems to be a distinction between main and dependent phrases only because of phonetic realization (or lack thereof) of recursion, but languages that lack copulae illustrate the principle more clearly.

2.3) There is also an argument that constructions like "there is" or intransitive words equivalent to "exist" are really just reflexive copulae that tie off structural loose ends. Therefore the statement above is equivalent to the proposotion "There is a hummingbird that is blue!" Potential investigation: ergative/absolutive languages with reflexive morphemes are known not to "cross systems", e.g. East Greenlandic, in which using ergative blocks use of the reflexive marker; you can only use one system at a time. So how do the reflexive copular constructions behave in ergative/absolutive languages that have both copulae and reflexivity?


3) DECLARATION: "I am going running at 3pm."

Extra-contential rhetorical tag: "I want you to accept as true the meaning of this proposition, and update your model of the world accordingly."

3.1) Although seemingly the most basic form of utterance, declarative propositions are not even close to the entirety of contentful utterances we make. Still they are zero-grade in all languages that do mark rhetorical intention.


4) YES/NO QUESTION: "Is he very tall?"

Extra-contential rhetorical tag: "I want you to reformulate this utterance as a proposition and then tell me your evaluation of its truth value."

4.1) Questions are usually marked, either by word-order changes, explicit particles (like Japanese -ka) or tone. English has few minimal pairs where tone makes a difference (e.g. permit) and such pairs are related, unlike full tone languages. Still, if the concept of minimal pairs is extended to rhetorical intention, tone is indeed explicitly encoded and certainly distinguishes minimal pairs. ("He is running for governor." "He is running for governor?" These sentences mean different things.)

4.2) Yes/no questions are therefore actually different kinds of utterances than those containing interrogative pronouns. In fact some languages do mark them differently. Latin marked verbs with the suffix -ne only in questions that did not contains interrogative pronouns.

4.3) In all cases that I know of, the verb dominates other parts of speech in taking on question particle - that is, if there's a verb in a sentence and a language has a question particle, the particle attaches to the verb. (Case in point, in Japanese, -ka typically goes on the verb but if a single-noun utterance requesting clarification, the question particle can go on the noun; "He's working in the city," one speaker says, and the other says "Takamatsu-ka,", i.e. "In Takamatsu?"

I have argued previously that verbs and adjectives are both first-order modifiers, but that some first-order modifiers can modify two nouns simultaneously (these are called transitive verbs.) In this view, nouns alone cannot create a proposition since there is no relationship stated between them without verbs. Therefore, it makes sense that the rhetorical marker would be placed on the verb that changes the utterance from a list into a proposition.


5) INTERROGATIVE-PRONOUN-CONTAINING QUESTION: "What is the best restaurant in Portland?"

Extra-contential rhetorical tag: "This is a proposition whose truth value cannot be evaluated since I have deliberately used a placeholder ('what'). I want listeners to reformulate the statement as a proposition but include information that can be plugged into the placeholder slot in such a way as to make the proposition true."

5.1) Interrogative-pronoun containing questions are also usually marked in some way (by word-order, tone, and/or explicit morphemes.)

5.2) Languages often have multiple interrogative pronouns for different types of nominal information, but never to my knowledge are there dedicated adjectival, verbal, or other interrogative words. Interrogative pronouns can be pressed into service for one-off service as verbs and even productively undergo morphosyntactic operations: (Imagine a woman has just been told her twelve year-old son was seen driving to school. "He was what-ing to school?" Nonetheless these kinds of operations on interrogative pronouns are never formalized.)


6) CONTINGENT DECLARATION: "If it rains today, you're on your own."

Extra-contential rhetorical tag: "I have explicitly marked off a proposition whose truth value is influenced by other propositions stated in close proximity and whose truth I may not be certain of, or by the way I obtained the information."

6.1) There are two sub-structures here: one is the typical if-then statement formulation for subordinate clauses we normally think of, but there is also the case of evidential markers most famous among Tupi-Guarani languages. Both systems are ways of explicitly marking the truth-weighting that the listener should give to the proposition so marked.


Though it doesn't merit a separate entry here, it's interesting that hypotheses aren't exactly questions, but they aren't exactly subordinate clauses either (though a hypothesis can be stated as both.) It's my suspicion that humans not engaged in research do not engage in extended hypotheticals - the propositions they are unsure about tend to be simple enough that their hypotheses are all contained in single clauses delineated by if-then markers. For most humans, thoughts complicated enough to require more than one sentence and which are of uncertain truth are merely deception, not hypotheses to be tested.

However if English does follow my humorous suggestion to develop an explicit hypothesis particle and a seventh utterance category, then I should re-state my earlier sentence as "A simple epistemological model of language is that all contentful utterances are commands, either directly to commit an action or to react to provided information, even if that's just for the listener to update his/her model of the world-ba."

The Verb Regularization Rate in English

"The half-life of an irregular verb scales as the square root of its usage frequency: a verb that is 100 times less frequent regularizes 10 times as fast." From Lieberman et al in Nature. An interesting question is to what other morphosyntactic rules this generalizes to, like noun plurals (and to what extent is it influenced by phonetic realization. My guess: not very much.) Pinker and many others knew qualitatively that the less a verb is used, the more likely it is to become regular in a given time period. Now we have the quantitative rule.